Ten things we believe.

Every fact you memorise is a node in a network. But fluency, expertise, and the ability to apply knowledge under pressure — these live in the edges: the connections between facts. A vocabulary list is a pile of nodes. The ability to form a sentence you've never seen before requires edges.

Plan2Skill is built on this premise. We measure your edges, not your nodes. We test whether you can use knowledge, not merely recall it. This is not a pedagogical preference. It is the core technical commitment of everything we build.

Duolingo has published research showing that users who maintain a 1,800-day streak often plateau after year one. The streak continues. The learning does not. This is not a criticism of Duolingo — it is a description of what happens when engagement becomes the goal instead of the instrument.

We will not do this. We will not give you points for things you already know. We will not congratulate you for repeating what you've mastered. We will celebrate progress, not activity. If you haven't learned anything new in a session, the session was not a success — regardless of how long you stayed.

We use RPG mechanics: quests, XP, archetypes, equipment, streaks. We believe these are legitimate scaffolds for motivation when they are tied to real progress — when you earn XP for forming an edge you didn't have before, not for clicking through a screen you've already seen.

The difference between a scaffold and a slot machine is whether it serves the learner or the retention metric. We are committed to scaffolds. We are opposed to slot machines. Every gamification mechanic in Plan2Skill is reviewed against this question: does this serve the learner's actual progress, or does it serve our daily active user count?

We do not publish the formulas, weights, and thresholds behind CTLT. They are our alpha IP, and we will not pretend otherwise. But we do publish the five theorists whose work grounds ours: Vygotsky (Zone of Proximal Development), Hebb (co-activation), Collins & Loftus (spreading activation), Bjork (desirable difficulty), and Kapur (productive failure). Every named claim is traceable to published research. Every mechanism is describable in plain language.

If you want to understand what we build on, read The science behind Plan2Skill. If you want to stress-test our claims, write to research@plan2skill.com. We will respond.

ChatGPT is a conversation partner. Plan2Skill is an instrument. You can learn guitar by chatting with someone who plays well. But a structured practice routine — with feedback, spaced repetition, and progressive difficulty — will get you further, faster. We use LLMs to generate content and adapt to your level. We do not use them as the interface, because the interface is the pedagogy.

A chatbot that answers your questions gives you information. An instrument that tests your edges gives you knowledge. We are building the instrument.

We are in alpha with one domain live. CTLT is not validated by a peer-reviewed RCT. Our sample is ~50 learners. We publish open metrics monthly — including the numbers that go down. We do not know yet whether our edge-weighting algorithm produces better outcomes than spaced repetition alone. We believe it will. We are testing this belief, not asserting it.

Every claim on this site that goes beyond what we can currently measure is marked as a belief or hypothesis. We will update this page when our evidence changes.

Sessions on Plan2Skill are variable-length. We do not have a fixed "lesson" format. Some quests take 4 minutes. Some take 25. We will not pad a 4-minute quest to 8 minutes to improve session-length metrics. We will not interrupt a 25-minute flow state with a congratulations screen to improve "session completion" rates.

If you have 7 minutes, we will give you a meaningful 7-minute quest. If you have an hour and are in flow, we will not interrupt you. Your time is real. Our metrics are proxies. We will not sacrifice the real for the proxy.

Most learning apps are designed for children or for adults who learn like children: short sessions, cartoon mascots, binary feedback, and a heavy reliance on extrinsic rewards. This works for some adults. It does not work for many.

Adults with ADHD, adults returning to education after decades away, adults learning professional skills under time pressure — these learners need different scaffolds. They need sessions that respect their attention span, not fight it. They need feedback that treats them as intelligent, not as a child who needs praise for showing up. They need to see their own graph — to understand where they are, not just receive a score.

The average adult language learner spends £4,000 on courses and apps before achieving intermediate fluency. Most of that money is spent on products not designed for them. We are designing for them.

We do not publish testimonials. Testimonials are not evidence. They are selected anecdotes, and we have no interest in selected anecdotes — including positive ones about Plan2Skill.

Instead, we publish live user counts, retention curves, quest-completion rates, and edge-formation rates on /metrics. These are updated monthly. They include the numbers that go down. A retention curve that shows 40% D7 retention is more honest — and more useful — than a testimonial from a user who loved us.

If you are evaluating Plan2Skill for yourself or your team, use the metrics. If the metrics are not enough, write to us and ask for more.

This is what we are building: a personal cognitive map that grows with every quest, in every domain you choose to learn. Languages today. Mathematics, programming, music theory, law, science — domain by domain, as we validate CTLT in each one.

The map is yours. You can see it. You can interrogate it. You can identify the gaps and decide what to fill next. You are not a passive recipient of an algorithm's judgment about what you should learn. You are a learner with a map, choosing your own path through the territory.

That is what Plan2Skill is for. Edge by edge. In every domain you choose.